Sign In

SIMPRO

VTT subproject

Computational methods in mechanical engineering product development

Demand

The everlasting search for new ways to gain a competitive advantage in business drives product development in companies to balance the contradicting objectives:

  1. technically innovative and advanced products should be designed
  2. in a continuously shortening time and
  3. with decreasing overall costs.

Companies that can balance this multi-objective optimisation challenge and have the right product, at the right time, and do successful marketing have a great asset in the business. The computational approach has proven to be a great tool, especially in engineering and design, but also in other areas in business, for making a balanced compromise in the above-mentioned optimisation challenge. Computational methods, including modelling, simulation, and analysis, have become standard tools of research and product development in industry worldwide, and many companies in Finland have applied them already for several decades. In scientific research, the application of computational methods alongside traditional theoretical and experimental research has gained general acceptance and, in practice, the computational approach has enabled cost efficient and fast implementation of research.

Nowadays, there are hardly any doubts about the usefulness of computational methods in research and product development, but this does not mean they are applied efficiently and effectively. The computational approach in research and development is a complex concept and contains several separate fields and layers. Getting the most out of it requires adequate mastery of the fields and layers, and an optimal balance of the efforts and investments. Knowledge and know-how about the large concepts and information about the achievable advantages are the necessary building blocks for making strategic decisions in organisations about the main processes and investments, both in the competencies and in the infrastructure.

Solution

High-performance computing in mechanical engineering

Different products need to be designed utilising computation based methods, as making prototypes is often too slow and expensive. Further, products are often needs to be tailored for customer's needs, adding attraction to computation. To take fully advantage of computational methods, high-performance computing (HPC) is often required. HPC enables new type of multiphysical computation; it makes possible to find reasons for previously unknown behaviour and enable optimisations of product designs. Here, the main objective is to lower the step for design and research engineer's to utilise HPC in her/his everyday work. In the following, five different typical challenges of HPC are discussed.

Computation management systems

High-performance computing often requires utilisation of quite complicated hardware and software systems. There are so-called computation management software tools to help the user to use computation resources more efficient and easier. The user needs not to know the computation hardware, but only give some requirements for the hardware and the computation management system handles the rest. These systems aim to optimise the computational load balancing and resource availability for the users letting the user to focus on the actual challenge in hands. Massive computations are run efficiently in different use scenarios and hardware environments, such as workstations, networks of workstations, servers, clusters, and supercomputers. In SIMPRO VTT subproject, use experiences are covered for the selected tools: Grid Engine, SLURM, HTCondor, and Techila. The conclusion was that some DRMSs can deal with several use scenarios, but none of them can handle all the scenarios well, and thus there is multitude of different workable systems.

Scripting in high performance computing environment

Effective use of HPC environment often requires the utilisation of scripting languages. Modelling work typically includes solutions to a variety of models, e.g. with different mechanical dimensions or material parameters, making the modelling a tedious and slow procedure. With scripting, we can automate this kind of modelling and develop users own routines which can speed-up the modelling and analysis time enormously. In addition, scripting enables integration between different software applications in an efficient way. One of the most popular present scripting languages is Python. In SIMPRO VTT subproject, Python was successfully utilised in product design optimisation, structural analysis, and visualisation.

Multi-physics simulation for electric machine development

Computational research on electrical machines has recently become more attractive due to fast development of simulation software and increase in computational power. HPC has not been widely utilised before with electrical machine electromagnetic simulations, due to limitations of the commercial electromagnetic finite element method (FEM) software tools. For massively parallel environments, software products licensed under open source license are very attractive in electrical machine analysis. Presently, the only available open source options for 3D simulations are Elmer and GetDP. Both have limitations, but Elmer has developed hugely during the SIMPRO project, and presently even exceeds commercial counterparts in several features. Figure 1 shows an Elmer solution of the 3D electrical machine end-winding case. For optimisation, analytical electrical machine models are preferred. In SIMPRO, an analytical model implemented in MATLAB was used to represent an electrical machine in the single- and multi-objective optimisation processes.

Figure 1. Elmer solution of the 3D electrical machine end-winding case

Co-simulation and parallelisation in technical simulation

In certain solutions, the forces caused by a fluid flow can deform a structure while the deformation again changes the fluid flow. Such cases requires model for fluid-structure interaction (FSI). In FSI, two demanding computational analysis domains are coupled to simulate interaction between fluid flow and structural dynamics, e.g. hydrodynamics and structural phenomena of propulsion systems. The video below shows results of a case study, where interaction of a rotating propeller and a non-rotating cylinder located at the wake of the propeller was studied. So-called weak coupling was used, i.e. each problem is solved separately and during each time step some variables are exchanged between the two problems. The MpCCI tool (from Fraunhofer-Gesellschaft) handles the connection between the two commercial software applications (Fluent CFD software and Abaqus structural FEM tool) in use.

 

Large-scale visualisation and open source tools in technical computations

The amount of the data produced by the simulations is continuously increasing, making the post-processing and visualisation of the results a more challenging task. Open source post-processing tool ParaView can utilise client-served mode, where all the simulation data need not been transferred from computation server to user's local computer. Further, parallel processing can be utilised. The ParaView visualisation can be rendered with open source 3D graphics software Blender as photorealistic figures. For example, The figure on the right shows a detail of the tube frame of the RC car model used as a case study. The video below represents the Salome platform – OpenFOAM – ParaView – Blender tool chain and the air flow results rendered using Blender. Further, open source optimisation software Dakota were utilised in the case of a drag minimisation of the airfoil when the minimum lift is given as a constraint. The objective function was calculated using OpenFOAM. Due to expensiveness of the objective function evaluations and numerical error originated from the coarseness of the grid, a surrogate-based optimisation method was utilised. There the objective function is replaced by simpler surrogate function.

Benefits and use cases

High-performance computing enables e.g. new type of multiphysical computation and massive optimisation. However, it has challenges both in hardware and software tools. In SIMPRO project, quite many quality solutions were found for these challenges. The solutions are documented in reports to benefit industrial users, who can adopt straightforwardly the best practises.

Thoroughly, open source tools were found to be potential for co-utilisation or even replacement of commercial computation software tools in HPC. Presently numerical efficiency means high utilisation of massive parallel computation, where open source tools like OpenFOAM, Elmer, Dakota, and ParaView have their strong points. These benefits are not only due to lack of licensing fees, but also due to more built-in parallelisation solutions. However, the commercial tools are still very much needed and co-utilised with open source tools. Further, open source tools often produce own ecosystem with support and solution providers and possible cloud services. Hence, open source should be seen as possibility for Finnish computation related enterprise sector.

Task deliverables

  1. Keränen, J. et al.
    Computational resource management systems
    VTT Research Report VTT-R-01975-15, Espoo, 2015
    Link: http://www.vtt.fi/inf/julkaisut/muut/2015/VTT-R-01975-15.pdf
  2. Katajamäki, K.
    Improving product design efficiency by automating computational work
    Poster, SIMPRO-ScarFace Joint Seminar 2, August 25–25, 2015, Espoo
  3. Katajamäki, K.
    Scripting in high performance computing environment
    VTT Research Report VTT-R-03683-15, Espoo, 2015
    Link: http://www.vtt.fi/inf/julkaisut/muut/2015/VTT-R-03683-15.pdf 
  4. Keränen, J. et al.
    Efficient parallel 3D computation of electrical machines with Elmer
    Magnetics, IEEE Transactions on Magnetics, 51(3), pp.1–4, March 2015
    DOI: 10.1109/TMAG.2014.2356256
  5. Sindhya, K. et al.
    Design of a Permanent Magnet Synchronous Generator using Interactive Multiobjective Optimization
    IEEE Transactions on Industrial Electronics, PP(99), pp. 1–1, May 2017
    DOI: 10.1109/TIE.2017.2708038
  6. Keränen, J. et al.
    Multi-physics simulation for electric machine development
    VTT Research Report VTT-R-04618-15
    Link: http://www.vtt.fi/inf/julkaisut/muut/2015/VTT-R-04618-15.pdf
  7. Keränen, J. & Manninen, A.
    Multi-physics simulation for electric machine development
    Poster, SIMPRO-ScarFace Joint Seminar 2, August 25–25, 2015, Espoo
  8. Nieminen, V.
    Fluid-structure interaction simulation utilising MpCCI
    VTT Research Report VTT-R-01447-15, Espoo, 2015
    Link: http://www.vtt.fi/inf/julkaisut/muut/2015/VTT-R-01447-15.pdf
  9.  Karvinen, A.
    Surrogate-based optimization of airfoil using open source software
    Proceedings of the XII Finnish Mechanics Days, June 4–5, 2015, Tampere
    Link: http://rmseura.tkk.fi/smp_proceedings/SMP12_Proceedings.pdf
  10. Karvinen, A.
    Open source tools in technical photorealistic large-scale visualisation
    VTT Research Report VTT-R-04911-15
    Link: http://www.vtt.fi/inf/julkaisut/muut/2015/VTT-R-04911-15.pdf
  11. Karvinen, A.
    Open source tools in technical computation and visualisation
    Poster, SIMPRO-ScarFace Joint Seminar 2, August 25–25, 2015, Espoo

Optimisation, design studies and analyses

Parameter variation and optimisation methods applied with calculation models can provide valuable additional information related to behaviour of structures. Depending on a case, studied structures can be very complex and complete virtual models are computationally very extensive.

Parameter optimisation

Automatic and integrated calculation systems can give clear benefits in modifying of existing products or during design process of new products. Modern and flexible analysis software combined with effective computer hardware and connected together with sophisticated scripting languages makes it possible to develop completely automatic, effective and reliable calculation processes. These processes can then be integrated within optimisation algorithms thus enabling product optimisation and control of the inevitable uncertainty in model parameters. In this subtask, another study case was a diesel generating set based on Wärtsilä Vasa 4R32 inline four-cylinder four-stroke engine, presented in the figure below. In this case, overall vibration levels were controlled and mass of the common base frame was minimised.

Engine internal processes were simulated with GT-Power and GT-Suite software tools in order to obtain engine excitation forces as harmonic components for vibration analysis. General layout of the excitation calculation process is show in the figure below.

These engine excitations were used as loading for the structural FEM-model. The complete calculation process was fully automatized and controlled by selected input parameters. The selected input parameters are presented in the figure below, some of these being continuous and some discrete.

Nowadays typical calculation models may have millions or even tens of millions degrees of freedom. As the optimisation requires easily thousands of cases to be analysed, becomes the analysis time for an individual case critical. It is then necessary to decrease the model size. One option is to use approximated surrogate models. Another way is to use methods to reduce the model size by reducing the size of sub-models, which remain unchanged in optimisation, by introducing user defined superelements. In this study, three different algorithms were tested in a case were a strong resonance is close to the nominal operating point. The best optimisation result was obtained from evolutionary algorithm, when the mass of the base frame was reduced by 28 %. Progression of the optimisation process is presented in the animation below.

Topology optimisation

Conventionally, topology optimisation has been used separately, mainly for new concept studies or for the optimisation of the shape of a new component to be manufactured with certain processes, like casting or additive manufacturing (such as 3D printing). The aim of the study was to find out how design space variation affects the topology optimisation results and also to automate the complete process, where also the optimised structure is analysed and tested against used constraints. The utilized process is described in the figure below.

When topology optimisation process proceeds, material is removed from areas where it is not needed. The process is illustrated in the animation below.

 

Structure optimisation using discrete element method

Discrete element method (DEM) is powerful method for simulating, for example, bulk material behaviour. A particle damper is an attractive alternative damper concept for reducing mechanical vibration, since it can attenuate vibrations within a wide frequency range. In this study DEM was tested for simulating a particle damper behaviour integrated to optimisation process with the computational power of parallel execution. Based on the simulation, it was shown that the particle damper concept works. Dynamics of the damper simulation is presented in the animation below.

Acoustic simulation in HPC and cloud environment

An attractive option to perform acoustic simulations using cloud computing services was tested. Results were compared to conventional FEM based simulations. The acoustic simulation test case was performed with Kuava Oy’s software application Waveller Cloud. The case study, Valtra cabin T888M, was concentrated on the sound field distributions produced by an acoustic point source into the interior of a tractor cabin, and simulation of material utilisation in three inner roof elements. When comparing results from both simulations, some similarities were noticed however distinct differences were also detected as seen in the figure below (left: Actran results, right: Waveller Cloud results).

Benefits and use cases

One of the task objectives was to explore different optimisation tools and to test the efficiently of different optimisation algorithms available in these tools. As one of main results from the case studies is that the importance of understanding the actual optimisation case which is being studied. If the optimisation problems is not described with necessary understanding, it could produce erroneous interpretation of optimised case or even impossible to get any meaningful result. In addition, different types and levels of analysis models were introduced in order to illustrate the effect of the target model content on the results.

Task deliverables

  1. Avikainen, T.
    Guide Rail Optimization by Isight
    VTT Research Report, VTT-R-04402-15. Espoo, 2015
    Link: http://www.vtt.fi/inf/julkaisut/muut/2015/VTT-R-04402-15.pdf
  2. Wennerkoski, J.
    Guiderail Optimization by HEEDS
    VTT Research Report, VTT-R-04432-15. Espoo, 2015
    Link: http://www.vtt.fi/inf/julkaisut/muut/2015/VTT-R-04432-15.pdf
  3. Rahkola, P.
    Parameter optimisation using Heeds MDO and DAKOTA
    VTT Research Report VTT-R-00998-14, Espoo, 2014
    Link: http://www.vtt.fi/inf/julkaisut/muut/2014/VTT-R-00998-14.pdf
  4. Katajamäki, K.
    Optimization of diesel-generating set vibration levels with Dakota
    VTT Research Report, VTT-R-04967-15. Espoo, 2015
    Link: http://www.vtt.fi/inf/julkaisut/muut/2015/VTT-R-04967-15.pdf
  5. Komi, E. & Hämäläinen, J.
    Topology optimization in HPC environment
    VTT Research Report VTT-R-00527-15, Tampere, 2015
    Link: http://www.vtt.fi/inf/julkaisut/muut/2015/VTT-R-00527-15.pdf
  6. Komi, E.
    Topology optimization – Design tool for additive manufacturing
    Poster, SIMPRO-ScarFace Joint Seminar 2, August 25–25, 2015, Espoo
  7. Klinge, P.
    Optimisation of a particle damper with DEM
    VTT Research Report, VTT-R-00651-14. Espoo, 2014
    Link: http://www.vtt.fi/inf/julkaisut/muut/2014/VTT-R-00651-14.pdf
  8. Klinge, P.
    Optimisation of a particle damper with DEM
    Poster, SIMPRO-ScarFace Joint Seminar 2, August 25–25, 2015, Espoo
  9. Uosukainen, S. & Siponen, D.
    Acoustic simulation in HPC and cloud environment
    VTT Research Report, VTT-R-04076-15. Espoo, 2015
    Link: http://www.vtt.fi/inf/julkaisut/muut/2015/VTT-R-04076-15.pdf
  10. Uosukainen, S. & Siponen, D.
    Acoustic simulation in HPC and cloud environment
    Poster, SIMPRO-ScarFace Joint Seminar 2, August 25–25, 2015, Espoo

Requirements traceability in simulation driven development

The different results of an engineering project tend to be scattered to separate data repositories that store results from a single phase of the project such as requirements acquisition, system modelling or simulation. This makes it a challenge to maintain consistency between the various engineering artefacts as they evolve during the project. One of the objectives of the VTT subproject of the SIMPRO project was to find solutions to facilitate traceability of design artefacts and their verification and validation reports to the original requirements.

Fixing the data model for simulation based system development

The VTT subproject of the SIMPRO project provided data models for capturing the dependencies between the engineering artefacts and tracing the impacts of changes in the artefacts. The set of models were titled Systems Engineering Artefacts Model (SEAModel). SEAModel consists of the packages depicted in the figure below (Figure 2).

Figure 2. SEAModel. (From the Speciality Engineering packages only the Risk assessment package is provided.)

The Requirement and V&V package is the core part of the model relating to the traceability of the requirements. The Requirement and V&V model was transformed to a traceability information model (TIM). The produced TIM is shown in Figure 3.

Figure 3. Requirement and V&V TIM.

Tracing requirements, system models and system simulation

Application of SEAModel and the consequent TIM was demonstrated with a typical set of engineering tools. The demonstration involved an fictitious mobile elevator platform (see Figure 4), for which a customer had requested a new feature to move the platform horizontally in the sideways direction. The new feature request caused a stability hazard that was analysed with the help of CAD (SolidWorks) and simulation (Simulink) models. The analysis results were traced to the mechanical models and to the safety requirement using a requirements management software tool (Rational DOORS).

Figure 4. The demonstration case.

The demonstration showed that the presented data models can be implemented to provide traceability with a typical set of engineering tools and IT tools that provide a structured data repository, such as a relational database. Even though full integration of data from various engineering tools cannot be achieved with just trace links, arranging the traceability according to the presented models makes a significant step forward towards good management of shared engineering artefacts, and also towards model-based Systems Engineering (MBSE).

Transferring requirements

In a typical development project, there are subcontractors that have to work with a set of requirements that originate from the main contractor. To transfer the requirements, a file format called ReqIF exists. The ReqIF specification is maintained by Object Management Group (OMG) (ReqIF 2013). The goal of ReqIF is to provide a file format to exchange requirements between different organisations or departments of an organisation developing a system.

In the SIMPRO project, the ReqIF format was evaluated by implementing a ReqIF interface for the Simantics integration platform for modelling and simulation. ReqIF data could be imported to Simantics after a moderate integration software programming; the ReqIF import mechanism was made generic to accept any kinds of XML files that are specified by XSD schemas. However, to provide full requirements roundtrip, new features (such as version control) for Simantics were noticed to be required.

Benefits and use cases

The results of the work are especially useful for SMEs that would like to improve maintaining the consistency of their engineering work products without investing in a full-blown Product Life Cycle Management tool. Due to the fact that SEAModel provides a model for a structured repository of engineering artefacts, SMEs can use the model to make their first steps towards MBSE. Such a structured storage of data facilitates automatic or semi-automatic generation of documentation for the different purposes of the customers and regulators.

Task deliverables

  1. Alanen, J. & Valkama P.
    Systems Engineering Artefact model – SEAModel
    VTT Research Report VTT-R-06629-13, Tampere, 2013
    Link: http://www.vtt.fi/inf/julkaisut/muut/2013/VTT-R-06629-13.pdf
  2. Tikka, P.
    Tracing requirements in simulation oriented mechanical engineering
    Master's thesis, to be published
  3. Mätäsniemi T. & Alanen, J.
    Implementation of ReqIF for the Simantics platform
    VTT Research Report VTT-R-04722-15, Tampere, 2015
    Link: http://www.vtt.fi/inf/julkaisut/muut/2015/VTT-R-04722-15.pdf
  4. Alanen, J. & Tikka, P.
    Tracing artefacts according to SEAModel in simulation oriented product development
    Poster, SIMPRO-ScarFace Joint Seminar 2, August 25–25, 2015, Espoo
  5. Alanen, J. et al.
    Requirements traceability in simulation driven development
    VTT Technology 236, Espoo, 2015
    Link: http://www.vtt.fi/inf/pdf/technology/2015/T236.pdf

Modelling and results data management

Today industries are required to provide new products with innovative and powerful features at accelerated pace. The integration of functionalities from various disciplines is a major source of innovation. Therefore products with multiple technologies need to be designed, verified and validated against multitude of requirements before being mass-produced. The use of digital tools is considered to be an essential solution during the design process and it is increasingly becoming common. However, development of complex products with minimum or no physical prototypes requires multitude of digital tools resulting in higher complexity of the design processes.

State of the art in modelling and simulation

The current trend is going towards the simulation-based product development, and simulations have become central part of the design process. However, to realize the simulation-based product development process in practice, many technological challenges need to be overcome. The most obvious challenge is that the digital world is not the real world, and thus involves many uncertainties which have to be taken into account when designing and simulating the developed product. Additionally, increased complexity in product design exponentially increases the complexity of the design process and thus the demands on verification and validation. New methods and new digital tools are constantly under development in order to deal with complex simulations and for evaluating the simulation results. Decision making process has become the crucial aspect of a product development and design evolution. Relevant information including simulation results and uncertainties shall be available to make the right decision and to develop the system further. Therefore, flawless management of design and simulation data is essential when the requirements are validated in the digital world and it is currently a major research trend.

Simulation lifecycle and data management

Simulation Lifecycle Management (SLM) tools allow the seamless combination of design and simulation tools within the same formalised process can reduce the risk of design errors early in the design, and enable the utilisation of design data in later stages of product lifecycle. As shown in the figure below SLM mainly focus on the virtual world of the product lifecycle. It manages simulations in a narrow collaboration with the product lifecycle management (PLM) and integrates likewise the capabilities of the simulation data management (SDM). SLM offers a clearer visibility for simulation data, and enables engineers and scientists to collaborate on a same simulation platform.

SLM compliments SDM and PLM by associating behavioural simulation data and processes to the digital mock-ups. It is at the core of simulation-driven design where analyses become a fundamental part of product development. By managing the simulation process, SLM ensures: what will be done, when, by whom, and where results will be stored and archived. As a result SLM transforms simulation from a specialty operation to an enterprise product development process.

Utilisation of simulation over the product lifecycle

The literature review and the discussions with the industry reveal numerous challenges from the complexity of the product to its life span and the number of teams working around the globe. The SLM tools offer the possibility to ease the CAD/CAE data sharing among the teams since it includes PDM and PLM systems. It also offers an efficient way of searching, updating and re-using simulation data. Versioning, data traceability and decision making processes are very useful features for a large team spread worldwide and the collaborating platform is simply accessible using a web browser with centralised database.

Benefits and use cases

During the SIMPRO project, several industrially relevant case studies were conducted using two state of the art commercially available tools. The first case study consisted of performing a simulation process that involved a widely and frequently used CAD software applications and a knowledge management tool. This process was implemented in the ANSYS Workbench and ANSYS EKM.

 

The second case study consisted of studying the management of large simulation file of an engine block shown in the figure below. Such files are very common nowadays in the engineering field. An Abaqus ODB file of 1 Tera Bytes with simulation results was used for the case study in the Dassault Systèmes SIMULIA environment.

The third case study is focused on to test how Abaqus data can be stored in the ANSYS EKM data management system and extracted.

The case studies show that the modern SLM tools can be used by industry to implement an improved product development and lifecycle data management process. SLM compliments SDM and PLM by associating behavioural simulation data and processes to the digital model of the product. It can enable the simulation-driven design as the analysis becomes a fundamental part of product development. The use of SLM tools transforms the simulation from a special process to an enterprise product development process. It can be a powerful tool to manage the simulation processes: What will be done, when, by whom, where results will be stored and archived, etc.

The SLM plays an important role while enterprises need to deal with multidisciplinary products and simulations. Data, information and knowledge are centralised on the same collaboration platform. The traceability is performed and extracted data are linked to their original sources. The computing system is part of the platform to ensure the accessibility control of the entire design process. Simulation processes and assessment activities are formalised and enable to close the loop between the validation assessment and the requirements.

However, the case studies also pointed out that practical implementation of SLM solutions is yet far from seamless. The integration of proprietary design and simulation tools remains a major challenge and the implementation of processes requires significant expertise. Additionally, the price of the tools and continuous and high dependence on the SLM tool vendors is a significant bottleneck, particularly for SMEs.

Task deliverables

  1. Campos, J. et al
    Industrial open source solutions for product life cycle management
    DOI: 10.1080/23311916.2014.939737
    Link: http://cogentoa.tandfonline.com/doi/full/10.1080/23311916.2014.939737
  2. Sibois, R. & Muhammad, A.
    State of the art in modelling and simulation
    VTT Research Report VTT-R-00800-15, Tampere, 2015
    Link: http://www.vtt.fi/inf/julkaisut/muut/2015/VTT-R-00800-15.pdf
  3. Sibois, R. & Muhammad, A.
    Simulation Lifecycle and Data Management
    VTT Research Report VTT-R-02486-15, Tampere, 2015
    Link: http://www.vtt.fi/inf/julkaisut/muut/2015/VTT-R-02486-15.pdf
  4. Sibois, R. et al.
    Simulation Lifecycle and Data Management (Case Studies)
    VTT Research Report VTT-R-03500-15, Tampere, 2015
    Link: http://www.vtt.fi/inf/julkaisut/muut/2015/VTT-R-03500-15.pdf
  5. Muhammad, A. & Sibois, R.
    Simulation lifecycle management tools for simulation-based product lifecycle
    Poster, SIMPRO-ScarFace Joint Seminar 2, August 25–25, 2015, Espoo

Contacts